---
name: "cross_industry_transfer"
description: "Borrow a proven mechanism from an unrelated industry and make it work in yours: strip a practice down to the causal mechanism that makes it work, sweep unlike industries for the same mechanism under other names, score fit and the cost of the move, and guard against surface copies that borrow the look without the mechanism. Use when someone wants to import a practice from another field, asks what other industries do about a problem, is copying a famous company's method, or needs to judge whether an outside idea will survive the move."
---

# Cross-Industry Transfer

## Purpose

Most hard problems have already been solved, in a different industry, under a
different name, by people who would never be invited to your meetings. This
skill runs the move properly: abstract the target problem into a function,
find the mechanism that solves that function elsewhere, map it element by
element onto the target, score whether the move is worth its cost, and pilot
it with a falsifier written in advance.

The method rests on settled work. Gick and Holyoak showed that analogical
problem solving runs in three steps: noticing a source, mapping its parts onto
the target, and applying the mapping, and that comparing two analogues helps
people extract the shared schema instead of the surface story. Gentner's
structure-mapping theory explains why: an analogy holds when the same system
of relations holds across different objects, so relational and causal
structure is what transfers, not appearances. Enkel and Gassmann named the
management version cross-industry innovation, or creative imitation: the
transfer of technologies, processes, solutions, or working principles between
sectors. Research on analogical distance adds the trade-off this skill prices
in: farther sources tend to bring more novelty and less ready fit, nearer
sources tend to bring better fit and less novelty.

Two worked transfers, each carried from source practice to working adoption,
are in references/worked-transfers.md: a Formula One pit crew reshaping a
children's hospital handover, and the Toyota Production System reshaping care
delivery at Virginia Mason.

## Workflow

### 1. Frame the target problem as a function

State the problem with a baseline number before looking anywhere else: what
outcome is wrong, how often, what one failure costs, and which constraint
makes it hard (time, hands, attention, risk, volume). Then rewrite it as verb
phrases with no industry nouns, for example "hand a fragile thing to the next
team with nothing lost" or "catch a defect where it is made, before it
travels". This step matters because memory and search both retrieve by surface
similarity. The function phrases are the search keys that find far sources.

If there is no baseline, stop and measure first. A transfer judged against no
baseline can never be proven to work.

### 2. Strip a source practice down to its mechanism

For any source practice, work down five layers and write each one:

1. Practice: what they visibly do, in their own vocabulary.
2. Actions: the behaviours, stripped of brand and props.
3. Mechanism: one sentence, no source nouns, stating what causes the outcome.
   Form: "When X happens under condition Y, Z follows because..."
4. Preconditions: what the source gets for free that the mechanism quietly
   depends on: staffing, training, authority, spare capacity, incentives,
   regulation.
5. Boundary conditions: where the mechanism should stop working, stated in
   advance.

A practice you cannot state at layer 3 is not understood yet. Do not map it,
score it, or pitch it.

### 3. Sweep unlike industries for the same mechanism

Use the worksheet in references/sweep-worksheet.md. Sweep by function phrase,
not by industry name, and fill at least six candidate rows across three
distance bands: near (inside or beside the target industry), mid (shares the
constraint, not the product), and far (shares only the function). Require at
least two far rows. Analogy-mining research frames the ideal find as near
purpose with far mechanism: the same job, done a completely different way.

Good hunting grounds for most problems: aviation and air traffic, nuclear and
rail operations, manufacturing and logistics, professional kitchens, stage
and film crews, racing pit crews, emergency and military teams, software
incident response, hospitality service recovery, maintenance trades, and
sport officiating. Treat the list as a starting pool, not a limit.

For each row record the mechanism sentence and where you found it. A row
sourced only to a vendor page or a keynote anecdote is marked weak evidence
on sight.

### 4. Map structurally, and map what does not fit

For each shortlisted candidate, build the mapping table from the worksheet:
source element, the role it plays in the mechanism, and the target element
that plays the same role. Keep the mapping one-to-one. If one source element
wants to map to two target elements, the mechanism is probably two mechanisms;
split it.

Then write the negative mapping: at least three source elements or conditions
that do not transfer, with a reason for each. If you cannot find three, you
are looking at a fantasy version of the source. Go back to layer 4 and 5 work.

### 5. Score fit and the cost of the move

Score each shortlisted candidate with the bundled scorer:

python3 scripts/score_transfer.py score --label "kitchen pass to ticket triage" --structural 3 --evidence 2 --preconditions 2 --pilotability 3 --burden 1

python3 scripts/score_transfer.py batch candidates.csv

python3 scripts/score_transfer.py demo

Five inputs, each 0 to 3: structural match, causal evidence in the source,
preconditions already met, pilotability (cheap, safe, reversible), and
adaptation burden (scored in reverse). The total is out of 15, and the hard
gates run first: structural match below 2 rejects, zero causal evidence
rejects, zero preconditions met holds the candidate until a precondition is
built or borrowed, and a surface-only flag rejects the candidate as a fake
transfer regardless of total. Verdicts: 12 to 15 pilot now, 9 to 11 pilot
after a named adaptation, 6 to 8 investigate only, below 6 reject.

Argue about scores in the open. Record the score the team argued about most,
and why it settled where it did. A score nobody argued about was probably
copied from the pitch.

### 6. Run the fake-transfer checks

Run all seven checks in references/fake-transfer-checks.md on the leading
candidate: rename test, strip-the-look test, counterfactual test,
precondition audit, negative mapping count, incentive check, and falsifier
check. Watch for the five common fake patterns named there: prop copy, title
copy, ritual copy, metric copy, and founder copy. A failed check must be
answered in writing, not waved through.

### 7. Adapt, pilot, and decide

Translate, do not transplant. Keep the mechanism invariant and change surface
features freely: vocabulary, tools, room layout, job titles, and timing can
all move, as long as the causal sentence from step 2 still holds in the
target. Build or borrow missing preconditions before the pilot starts, or
name the pilot as a precondition test and judge it as one.

Pilot on the smallest scope that still produces the real outcome: one team,
one shift, one queue. Set the duration, the primary metric tied to the step 1
baseline, and the falsifier in advance. Rehearse where the source rehearses.
Debrief after each run and feed errors into the next version of the sequence,
the way both worked examples did.

Close with one of three decisions, in writing: adopt (the metric moved, the
falsifier did not fire, the preconditions hold on a bad day), adapt and
re-pilot (the mechanism shows life, one named weakness explains the gap), or
drop (the falsifier fired, or the cost of the missing preconditions exceeds
the gain). File the transfer record either way, including the negative
mapping, so the next team does not re-import the same failure.

## Output Contract

Deliver a transfer brief with these parts, in this order:

1. Target problem: baseline number, constraint, and function phrases.
2. Candidate table: at least six swept candidates with distance band,
   mechanism sentence, and source for each.
3. Shortlist mapping: for each finalist, the mapping table and at least three
   negative mappings.
4. Scores: scorer output for each finalist, plus the most-argued score and
   its reasoning.
5. Fake-transfer checks: the seven checks on the leading candidate, each
   marked pass or answered.
6. Pilot card: mechanism invariant, preconditions met and missing, scope,
   duration, primary metric, falsifier, reversal plan, owner, review date.
7. Decision record after the pilot: adopt, adapt, or drop, with the measured
   result against the baseline.

Keep the brief itself short enough to read in one sitting. Put the full
worksheet, CSV, and scorer output in supporting files.

## Operating Rules

1. Mechanism before admiration. No candidate moves forward without a
   one-sentence causal mechanism that contains no source nouns, no brand, and
   no famous name.
2. Surface similarity is a retrieval clue, never evidence of fit. Fit is
   argued from relational structure, preconditions, and causal evidence only.
3. Every factual claim about a source practice needs a checkable source:
   official documentation, a study, a reputable report, or direct observation.
   A success story told by the source itself is a lead, not evidence, and it
   caps the evidence score at 1.
4. Negative mapping is mandatory. A transfer plan with nothing that fails to
   map has not been thought through, and it does not ship.
5. Borrow only what is public. Use published, taught, or openly shared
   practices. Do not solicit another organisation's confidential material, and
   check patents, licences, and trademarks before adopting a named method
   commercially.
6. Regulated and safety-critical targets need a domain expert from the target
   industry to sign the mapping and the pilot card. A good analogy does not
   override clinical, aviation, legal, or engineering judgement in its own
   field.
7. The pilot must be reversible, and the reversal plan is written before the
   pilot starts. If a pilot cannot be undone, it is a launch decision, and it
   leaves this skill for the owner's normal approval route.
8. Report results against the baseline, including when the transfer failed.
   A dropped transfer with a clean falsifier is a good outcome. A vague
   success with no numbers is not an outcome at all.

## Sources

- Gick and Holyoak (1980, 1983) on analogical problem solving, summarised at
  https://nmoer.pressbooks.pub/cognitivepsychology/chapter/problem-solving-by-analogy/
- Gentner (1983), structure-mapping theory, Cognitive Science 7(2), and
  Gentner and Smith (2013), reviewed at
  https://link.springer.com/article/10.1007/s13164-018-0419-y
- Enkel and Gassmann (2010), Creative imitation: exploring the case of
  cross-industry innovation, R and D Management 40(3), defined and cited at
  https://link.springer.com/chapter/10.1007/978-3-030-61295-5_9
- Brunswicker and Hutschek (2010), Crossing horizons, International Journal
  of Innovation Management 14(4), at
  https://www.worldscientific.com/doi/10.1142/S1363919610002829
- Hope et al. (2017), Accelerating innovation through analogy mining, KDD
  2017, summarised at https://blog.acolyer.org/2017/09/14/accelerating-innovation-through-analogy-mining/
- Great Ormond Street and Ferrari handover reports at
  https://www.qualitydigest.com/inside/innovation-column/outsourcing-innovation-022514.html
- Virginia Mason and Toyota reports at
  https://todayshospitalist.com/to-make-its-processes-lean-this-hospital-took-a-page-from-the-japanese-car-industry/
  and https://pmc.ncbi.nlm.nih.gov/articles/PMC3678835/